Fig. 5
From: Gene clusters based on OLIG2 and CD276 could distinguish molecular profiling in glioblastoma
Validation of gene clusters in GSE84010 and Gravandeel’s GBM datasets. Heatmap shows good classifying ability of gene clusters in two independent datasets (A, D). PCA analysis reveals gene clusters could distinguish mesenchymal and proneural subtypes perfectly, but less distinguishable from classic and neural subtypes (B, E). ROC of random forest classification model reveals good efficacy (C, F) (AUC = 0.816 in the GSE84010 dataset, AUC = 0.820 in Gravandeel’s dataset)